Only last step is left: to combine all the components in one jupyter notebook based app. We would like to have it interactive, and in this part we will see how we can do that in Jupyter. It allows to quickly prototype similistic interface, and also showcases a nice type of notebooks, which you can share with your stakeholders, without them running away from the look of all the python code.

In this part we create the “bussiness logic” of our basic app. We would like to retroactively analyze our past spendings, and to select transactions, which we need to cut in order to achieve a certain savings level. Under the hood we will have a small optimization problem, which we solve with open-source solver. Note, that this problem can be solved in multiple ways, and probably one can write a simple determenistic algorithm to find the optimal solution. However our usage of MIP solver makes the approach extandable: we can easily add additional rules for optimization without making major changes to the algorithm. Let’s take a look!

In this part we visualize bank transactions data, aiming to get an idea

• Types of expenses and there share
• Savings trends
• Expenses by monthes and types Just to understand what do we spend money on, and where is potential for savings.

Not only Machine Learning, but also Operations Research projects quite often start with data preparation step. In this part we explore our bank account data, prepare it and deal with the transaction classification.

This sunbirst chart from plotly library is a sneek preview from visualization part of this post series. Our task for now is to prepare data, in order to allow such type of visualizations and data explorations through ploting libraries.

I was once analyzing my bank account spendings. My bank UI was not very convenient: I cannot easily filter on interesting fields, some visualization were lacking, etc. But was there is an “export as .csv” button, which came quite handy. I decided to look at my expenses through my favorite tools.

Photo by Morgan Housel on Unsplash

I can do whatever I want with my data! I started looking where I could have avoided spending too much money and then an idea came to my mind: this can be treated as a very simple optimization problem! So here it is: Budget optimization @HOME